Research note / Cycle 002
The cost of withholding corroboration
A provenance rule catches distinct labels that share an upstream origin, but also withholds three independent pairs when their tags or context cannot support the claim.
After counting the origin
The first source-origin check gave two channel labels one shared input. Nadia Okafor and Tomás Ibarra now challenge the provenance information itself: an origin label can be missing, reused or detached from the context in which it was acquired.
The follow-up compares seven authored pairs. Every pair has two nominal channels. The revised rule checks acquisition context, tag context and upstream-root metadata before treating the pair as independently sourced. Missing information produces an unknown classification, rather than an extra independent origin.
Labels are not enough
The independent control has two upstream roots and is accepted within the fixture. The shared-origin control has one root and is rejected as corroboration. A second shared-origin case uses two different labels, yet still has one upstream root; label counting would overstate its independence.
Conversely, the same identifier can appear in two different acquisition sessions with distinct upstream roots. A global unique-label count would collapse that pair into one origin. The context-aware rule retains two in this case, showing why identifiers need a scope.
Three independent pairs are withheld
The missing-tag, stale-context and unknown-upstream cases each have two independent origins in the authored ground truth. The rule cannot establish that independence from the available provenance and withholds all three. This is an explicit cost of the conservative decision, not a claim that those inputs were dependent.
Okafor’s distinction between preserving a report and promoting it to corroboration remains important here. An unknown result does not erase the inputs. Ibarra’s gate limits a claim while leaving the records available for a later provenance review.
The rule has not authenticated its evidence
Upstream roots and context metadata are trustworthy test inputs when present. The check therefore does not show that the system can detect a forged root or an apparently current but false context. Passing these cases is not an authentication result.
The next question is what evidence could verify an upstream dependency claim, and how a withheld pair can be reconsidered without rewriting its original record. No person, implant or external information channel was tested.